In the last article, polling errors that defined the 2016 election and were secretly apparent in the 2020 election were examined and understood. In this article, remediations to the aforementioned methodological mistakes and possible future errors will be examined. On this, hours before a Presidential debate that shows a deadlocked race and a fight in the electoral college that’ll come down to the narrowest of margins.
Where Pollsters Have Improved
The first, and greatest change, is in what responses are used. Following a study by NYTimes/Siena, one of the best national pollsters, had they counted responses that erred along the lines of cursing at the pollster and stating they are voting for Trump, nearly half the polling error apparent in 2020 would be accounted for. This goes against the 2016 “shy Trump voter” theory (the idea that people were voting for Trump but would not admit it publicly) and shows that many Trump supporters are open about it, instead, they follow Trump’s words on not trusting pollsters.
Furthermore, on the 2016 issue of not accounting for education when weighing poll results, this is an easy fix that was implemented in 2020 and will continue to be implemented today.
Most importantly, are the changes done to weighting and sampling. Unlike before when landlines dominated how pollsters got data, pollsters are now using a wide range of polling methods; texting, for example, has been used far more often, leading to a more representative sample. The greatest change however is asking respondents who they remember voting for (giving the name “recall” questions that occur now in polling). This is critical as it ensures that people who voted for Trump last election are being accurately polled and weighed unlike in 2020 where Democrat voters were oversampled.
Where Pollsters May Fail
Despite all these changes, there is no doubt going to be some error this election, to what extent however we do not know.
Some clarity is offered in the 2022 midterm election, landmarked by a major abortion ruling. An election that at face value had very successful polling: pollsters predicted a slight Republican edge that was materialized in the popular vote. On the national scale, many of the methodological improvements that pollsters discovered following the 2020 election worked. However, this was on the national scale not state level.
Many might remember a so-called “red wave” that seemingly never materialized. This rested on one fact, Democrats overperformed in senate elections in the states that mattered. Although the popular vote was in fact a slight Republican lean, in safe blue states like New York, Republicans managed surprise victories and well overperformed in the 2020 election. Yet in swing states—Pennsylvania, Wisconsin, Michigan, Georgia, Arizona, Nevada— Democrats overperformed polling and pulled off what some might call surprise wins and achieving 51 senators (including the two independents who caucus with Democrats such as Bernie Sanders). Thus, although nationally pollsters were accurate, in the states that mattered, they were not.
Knowing this, many have theorized that Harris would need fewer votes nationally to win the electoral college as the states that matter, the swing states, are what decide elections. If the polling errors that occurred in the 2022 midterm happen this year, Harris will win handily and coast to victory as the first woman president; if pollsters are still unable to poll Trump voters properly, a surprise victory will push Trump to become the 45th and 47th United States President.
Ultimately, no one will know the answer until enough ballots are counted on election day (and perhaps the next couple of days later if we’re in a nail-biter election). Nonetheless, people can make predictions and be confident that it is unlikely for such a polling miss that had happened in the prior two elections to occur again.
Leave a Reply